October 20, 2025Open Access

H2HTalk: Evaluating Large Language Models as Emotional Companion

Key Points

H2HTalk sets the first benchmark for assessing emotional intelligence in LLMs, aiming for authentic companionship.
Evaluation of 50 LLMs indicates deficiencies in long-horizon planning and implicit user needs during interactions.
The benchmark includes 4,650 diverse scenarios, significantly enhancing the available datasets for model evaluation.
Incorporating attachment theory principles with a Secure Attachment Persona module supports safer emotional interactions.

Abstract

As digital emotional support needs grow, Large Language Model companions offer promising authentic, always-available empathy, though rigorous evaluation lags behind model advancement. We present Heart-to-Heart Talk (H2HTalk), a benchmark assessing companions across personality development and empathetic interaction, balancing emotional intelligence with linguistic fluency. H2HTalk features 4,650 curated scenarios spanning dialogue, recollection, and itinerary planning that mirror real-world support conversations, substantially exceeding previous datasets in scale and diversity. We incorporate a Secure Attachment Persona (SAP) module implementing attachment-theory principles for safer interactions. Benchmarking 50 LLMs with our unified protocol reveals that long-horizon planning and memory retention remain key challenges, with models struggling when user needs are implicit or evolve mid-conversation. H2HTalk establishes the first comprehensive benchmark for emotionally intelligent companions. We release all materials to advance development of LLMs capable of providing meaningful and safe psychological support.

H2HTalk: Evaluating Large Language Models as Emotional Companion

Key Points

Abstract

Cite This Study

Also Consider

Also Consider